List your company

Reasoning Question Answering on MuSiQue (test)

26.8p@1GARL
13.07216.63620.223.764Apr 28, 2026

Evaluation results

MethodLinks
GARLq=0.75, Setup=cold-start (no prompts)2026.0426.858.934.8
GARLq=1, Setup=cold-start (no prompts)2026.0421.658.132.5
GRPOSetup=warm-start (prompted)2026.0413.637.315.8